Clustering Key Features

Heatmap

To get a broader overview of the similarities between several tracks, I created a heatmap with seven characteristics I thought would show the most interesting differences between tracks and groups of tracks: liveness, tempo, instrumentalness, loudness, energy, valence and acousticness. I used the 20 longest tracks from the playlist ‘’Iconic Soundtracks’’, because the full playlist was too much to handle for my pc. And longer tracks have more variation in them, so it makes the analysis even more interesting.

What strikes already from the first glance that there are two outliers on two different characteristics: ‘’With my own eyes’’ scores very high on liveness and ‘’Flight to Neverland’’ very high on valence. However this is all relative, the liveness score is just 0,21, and this track has no audible audience, so in this sence the yellow colour can be a bit misleading. The same goes for the high valence, which is 0.35 and could be much higher. But in this corpus, they are the outlier on these factors.

Another interesting analysis can be made by how the characteristics are grouped. Tempo and instrumentalness are grouped together, which implies that on general, tracks with low intrumentalness have also a lower tempo and vice versa. But not all classifications regarding instrumentalness are as accurate, since many soundtracks consist of some choir and sometimes the API picks it up, but sometimes it doesn’t. So it doesn’t say much in general. Loudness and energy are correlated in the same way according to this heatmap and it makes sense: louder soundtracks usually convey more energy, since they consist of more bombastic elements and energetic choirs. Especially when you listen to the track ‘’Raider’s March’’ which has an energy score of 0.35.

As you can see, there are many clusterings for the tracks and they make sense. For example, the second and third row from below show the connection between ‘’Pure Spirits of the Forest’’ and ‘’Blade Runner Blues’. You can see they are very similar on all of the characteristics. But when you listen to these tracks, they still show some differences. Pure Spirits of the Forests isa piece which builds up very slowly with calm and peaceful elements and has an energetic part at the end. Blade Runner Blues has the same calmness but has a way fuller sound throughout the entire track. But here the same problem as before can be seenm the colours are not always accurate: the loudness is correct, and tempo show the same colour which makes sense, since they have a similar tempo. Liveness is not correct since they both have no sound of an audience in the recording.

Tempi Hans Zimmer & John Williams

Tempi Hans Zimmer & John Williams

Here you can see a histogram in which the different tempi of the soundtracks by Hans Zimmer and John Williams are compared. It is clear that Hans Zimmer uses a larger variety of tempi compared to John Williams, and Hans Zimmer has more tracks with a larger tempo. This concentration in one aspect was also seen before in the analysis of keys, which showed that Hans Zimmer used a larger variety of keys and John Williams focused on a select group of keys to be used in his tracks.

Tempi Iconic Soundtracks

Using the playlist Iconic Soundtracks, I created a histogram in which you can see the tempi of the soundtracks. Here you can see that in general, the tempo lies between 60 and 150 bpm with some outliers. I wanted to see if the outliers’ tempo was actually correct or a mistake and that perhaps the BPM was doubled or halved and for this I listened the songs and tapped with a BPM counter.

For the first 2 outliers on the lower end, I realized it was halved so in fact the speed was around 100 bpm. I think the mistake was made because the instrumentation was quiet. For the third one (Nino Rota - The Godfather Waltz), the bpm was correct, therefore I created a cyclic tempogram for it to find out any interesting features of this outlier. It is clear that this tempopgram doesn’t show the real bpm, but a lot of activity is seen around 100 BPM, especially between 30 and and 70 seconds. If you assume that the BPM is doubled in the entire tempogram, it is correct. After this mark the song gets even more quiet with only a sad violin, it is clear that not much information is picked up here, the same as in the beginning of the song. But after around 110 seconds, a piano starts to sound with a faster pace, this makes clear why here there is a yellow vertical line, the algorithm doesn’t know what the exact tempo is anymore.

It was also interesting to analyze an outlier on the other end. The four fastest ones, with the fastest almost at 200 bpm were clearly doubled and one even tripled. The real outlier is Hans Zimmer - Discombobulate, the fast paced soundtrack for the movie Sherlock Holmes. When you listen to it, it is easy to follow the speed, since the drums and other instruments clearly map out the rhythm. This is shown clearly in the cyclic tempogram, which correctly outlines the speed of 142 bpm.

The Keys in the music from Hans Zimmer and John Williams

Keys

In this histogram you can see the keys that are used in the soundtracks by Hans Zimmer and John Williams. It is clear that the most used keys by Hans Zimmer are D and G and by John Williams C is most often used. What you can also see from this histogram is that Hans Zimmer uses a wide variety of keys in his soundtracks, having at least used every key once. John Williams didn’t use every key and his use of keys is more concentrated.

Two Chordograms “Hans Zimmer - I’m not a Hero” & “John Williams - Duel of the Fates”

Here I dive deeper on two famous works from Hans Zimmer and John Williams, to see if some differences can be found using a chordogram for both tracks. Firstly, an analysis the track I’m not a Hero, the soundtrack for the movie Batman: The Dark Knight. What is clear already from a first glance, is that Hans Zimmer uses a concentration of a couple of the same chords throughout the entire track, but that there is still a variety of other chords on top of that to keep the track dynamic and unique. From the chordogram of Duel of the Fates (the soundtrack for Star Wars: The Phantom Menace), it can be seen that John Williams uses a much wider array of chords throughout the entire track. When you listen to this track it also makes sense, it is a more bombastic track, consisting of a powerful choir and a large amount of instruments. As two tracks are not enough to draw large conclusions from this comparison, it can still be said that in this case Hans Zimmer uses a more subtle and minimalistic approach regarding chords and that John Williams uses a larger melodic and harmonic soundscape at once. Hereafter, you can see a larger analysis of multiple tracks to build further on this comparison to see if this is the case for more tracks, using other ways of analysis as well.

Central point in scatterplot: Yann Tiersen - Comptine d`un autre ete-l’apres-midi

Amelie Full

As interesting it is to analyze outliers, it is also relevant to analyze a central point in the corpus, a soundtrack which can be classified as averege. I decide to choose a point on the intrumentalness/energy scatterplot, with instrumentalness around 0.90 and energy around 0.2. A song which fit this well is Yann Tiersen - Comptine d`un autre ete-l’apres-midi, the famous piano soundtrack of the French movie Le Fabuleux Destin d’Amélie Poulain. For this I created two cepstrograms, the first one based on bars and the second on sections. At the first cepstrogram you can clearly see that there is a steady basis in the song which flows during the entire song, with a very high magnitude at the start and maintaining this magnitude throughout the entire song, only losing some strength at the end. Furthermore, it shows some extra activity throughout the song in some sections indicated by the light blue colors. At the second cepstrogram it can be seen that there are seven main sections, of which the first has the strongest magnitude at the root note. After the start of the song, more energy starts to appear in the higher sections. Then there is a moment of calmness when the color turns darker green, followed by a short section of lighter green, and darker green again. This is a good representation of a soundtrack, which is usually divided in several sections which are not evenly divided like in pop songs.

Self Similarity Matrices: Hans Zimmer I’m not a hero

Self Similarity Matrices: Timbre & Chroma

Here you can see two self similarity matrices for the soundtrack I’m not a hero created by Hans Zimmer. I chose this one for this analysis, because it is a very famous soundtrack, and personally one of my favourites. I was curious to find out if the originality in the music is somehow reflected in these matrices.

In the first one, regarding timbre, it can be seen that the beginning is much like the rest of the song and that later on, there are new elements introduced. and moments of uniqueness and similarity are alternated. This makes sense, it starts with a slow buildup with many elements that are there throughout the song and which are later accompanied with more instruments to make it more cinematic and powerful. Throughout the song, slower moments are alternated by more powerful moments. The bright yellow line indicates a loud crash accompanied by brass sounds rising in intensity. This is a unique moment in the song.

In the second matrix regarding chroma (pitches), a different pattern can be seen. It is indicated that around the first half of the song, many unique pitches are heard and through the second half, mostly pitches that are common in the song are heard. This can be explained by the fact that the beginning of the track consists of several high pitched percussion instruments which are not heard much more in the song. The majority of this song consists of either very quiet or bombastic moments and these moments occur mostly during the second half of the song.

A Global Analysis of Famous Soundtracks

Instrumentalness vs. Energy

Here you can see a scatterplot from the playlist ‘’Iconic Soundtracks’’, which also consists soundtracks of various other composers and here I compared instrumentalness and energy. Here it is clearly shown again that most tracks score high on instrumentalness and have quite low energy levels. Again, certain outliers on both levels still exist. What I think is interesting from these analyses is that it shows that even two composers show quite some variety in the characteristics of their tracks, but that roughly speaking there are also similarities. I am interested to further investigate comparisons from other big composers and from main famous soundtracks in general, to find out more interesting differences and similarities. And I am also interested to further analyse specific outliers, to find out what is so different about them.

Comparing Hans Zimmer & John Williams

Hans Zimmer vs. John Williams

These two graphs show the main differences of the music by Hans Zimmer and John Williams. First, you see an analysis of characteristic track features of famous tracks created by them. For both of them I used the ‘’This is’’ playlist from Spotify. I made a visualization which shows if a song is major or minor, and showed the instrumentalness and valence levels for their tracks. The size of the circles show the energy level of the tracks. What you can clearly see already is that from both composers, most songs show high instrumentalness and low valence and this is most evident for Hans Zimmer. And he also has some tracks which score low in instrumentalness and low on valence, which I called ‘’sombre vocalization’‘. But he also has some outliers which are high on valence, on a level that the outliers of John Williams don’t reach. This is what I called’‘happy instrumentation’’. It is also clear that the majority is written in minor and that a large level of songs has an average energy level.

The Outlier Kawai - Utai IV: Reawakening

Chromogram of Kenji Kawai - Utai IV: Reawakening

I thought it was interesting to create a chromogram of one specific outlier to see if something extraordinary can be found about this soundtrack. This is the track Kenji Kawai- Utai IV: Reawakening. It is represented by the the dot all the way in the upper left of the scatter plot, which has the lowest instrumentalness and the highest energy. First it is noticeable that there is a strong emphasis on the D note in the beginning of the track. Furthermore, the highest magnitude can be found in the C and C#/Db notes throughout the song. And there are two main moments in which the notes are clearly rising, and three main moments where they are descending. You can hear it in the song when a strong emphasis is put on a choir of which the vocals are rising. And somewhere in the middle the highest point is reached, on a B note. What could make this different from other soundtracks is the strong emphasis on vocalization. But the fact that it builds up clearly and has a steady ongoing flow of the same notes is not so uncommon in soundtracks. To find out more about this, I have made different chromograms of soundtracks, as you can see in the other parts in this portfolio.

Concluding the Project